lgcre 0.0.3

Local Graph Context-based Recommende Engine. This is a library to build peronalized POI (point of intereset) recommender engines on the device. Specifiy a graph of related tags or categories, and whic [...]

LGCRE (Local Graph Context-based Recommender Engine)

This is a library that implements a recommender engine for points of interest (pois) on the device. The library is meant to serve recommendations to a "person", the device owner, based on building a graph of pois and categories around the person. Then given a set of pois and their categories, the

Getting Started

Look at examples/lib/main.dart on how to use the plugin.

// LGCRE Initialization

Map<String,String> options = {
  'lgcre.licenseid' : '<license_id>',    //<- The license id (Required)
  'lgcre.clientid' : '<client_id>',      //<- The client id (Required)
  'lgcre.loglevel' : 'fine',             //<- The log level (fine, off, debug, severe, config, warning)
  'lgcre.maxmem' : '64'                  //<- The max amount of memory to use
};
Lgcre.init(options);

// LGCRE reset data
Lgcre.reset();

// LGCRE add data
// Each data key (pois, categories, link_categories,
// etc...) can be null if no data needs to be added from that type. In
// this example we add data for all types just as an example

Map<String,dynamic> data = new Map<String, dynamic>();
// Add pois to the recommender. For instance, when we know the person has
// visited one poi and we want to represent this by means of a link
// between the person and the poi (see below)
data['pois'] = ["poi1", "poi2", "poi3", "poi4"];

// Add categories to the recommender. For instance, because we know the
// person is interested in a particular category or because we know that a poi
// the user has visited, has a category (see below)
data['categories'] = ["cat1", "cat2", "cat3", "cat4"];

// Add links between categories to the recommender. For instance, when 
// we know that two categories are related. Both categories must have been
// previously added before through the categories field, either in this
// call or in a previous call to 'addData' (see above). This field is used
// to build a 'graph of categories'. It is important to make this graf as
// connected as possible since this may greatly affect quality of the results
data['links_categories'] =  [ 
{'category1' : 'cat1', 
  'category2' : 'cat2'}, 

{'category1' : 'cat1', 
  'category2' : 'cat3'},

{'category1' : 'cat2', 
  'category2' : 'cat4'},
];

// Add links between pois and categories to the recommender. When we know
// that a poi is related in some way to a category. Both pois and
// categories in the links must have been added previously through 'pois'
// and 'categories' field, either on this call or in a previous call to
// 'addData' (see above).
data['links_poi_category'] = [ 
{'poi' : 'poi1', 
  'category' : 'cat1'}, 
{'poi' : 'poi2', 
  'category' : 'cat2'}, 
{'poi' : 'poi3', 
  'category' : 'cat3'}, 
];

// Add data to the person. 'pois' are pois related to the user, for
// instance, because we know he has liked or visited it. The pois in the
// list must have already been added through the "pois" field, either in
// this call or a previous call to 'addData' (see above). Categories are the 
// categories the person is interested in. These categories must be added inthis
// through the 'categories' field or in a previous call to 'addData' (see above)
data['person'] = {
  'pois' : ['poi1', 'poi2'],
  'categories' : ['cat1', 'cat2'],
};

Lgcre.addData(data);


// LGCRE Run a recommendation query 
// To execute a query, build a map between pois and lists of categories as
// follows
Map<String,dynamic> query = new Map<String,dynamic>();
query['pois'] = [
{ 'poi' : 'poi1', 
  'categories' : ['cat1, cat2']},
{ 'poi' : 'poi2', 
  'categories' : ['cat1, cat3']},
{ 'poi' : 'poi3', 
  'categories' : ['cat2, cat4']},
{ 'poi' : 'poi4', 
  'categories' : ['cat3, cat4']},
];

// The method run
List<ResultEntry>? result = await Lgcre.run(query);
if(result != null)
{
  for (ResultEntry entry in result)
  {
    print(entry.poi+" "+entry.score.toString()+"\n");
  }
}


// LGCRE: Releasing resources 
Lgcre.release();

Android

You will need to include the proguard rules to your app project. Add the file example/android/app/proguard-rules.pro to your app project, and include the following configuartion to

buildTypes {
  release {
    proguardFiles getDefaultProguardFile('proguard-android.txt'), 'proguard-rules.pro'
  }

Alternatively, you can disable proguard from your project by adding to

buildTypes {
  release {
    shrinkResources false
    minifyEnabled false
    useProguard false
  }
  ```